image guidance and registration – …0 20 40 60 80 0 0.2 0.4 0.6 0.8 1 dose (gy) volume (%)...
TRANSCRIPT
IMAGE GUIDANCE AND REGISTRATION –PARTICULAR CHALLENGES IN CASE OF RE-IRRADIATION
MARTIN F. FAST
ACKNOWLEDGEMENTS / SLIDE CREDITS
2
Tessa van de Lindt (NKI-AvL)
Jan-Jakob Sonke(NKI-AvL)
Uulke van der Heide (NKI-AvL)
Simon van Kranen (NKI-AvL)
CONFLICTS OF INTEREST
• Financial and technical support from Elekta AB under research agreements
• NKI-AvL is part of the Elekta Atlantic MR-linac Research Consortium
3https://mrrt.elekta.com/
OUTLINE
THE FUTURE IS NOW – MRI-GUIDED RADIOTHERAPY
THE WORKING HORSE – CBCT-GUIDED RADIOTHERAPY
DOSE ACCUMULATION – READY FOR CLINICAL USE?
IMAGE REGISTRATION – CHALLENGES AND OPPORTUNITIES
4
ELEKTA UNITY MR-LINAC
• 1.5 T Philips magnet
• 7 MV linac
• 22x56 cm2 treatment field
• Real-time cine-MR imaging
• MLC based on Agility design
Photo courtesy: Elekta AB 5
ELEKTA UNITY MR-LINAC AT NKI-AVL
INSTALLATION• May 2016
INITIAL RESEARCH PHASE• Phantom measurements• Volunteer imaging• Patient volunteer imaging
FINAL UPGRADE• June-August 2018
6
ELEKTA UNITY MR-LINAC AT NKI-AVL
First patient treated TODAY (September 6th, 2018)
7
Clinical focus• Prostate cancer• Rectal cancer• Oligometastases (especially liver)
Technical focus• EPID dosimetry• Practical 4D solutions• Quantitative imaging
MR-LINAC AT NKI
8
NKI’S LIVER RT WORKFLOW FOR THE MR-LINAC
PRE-TREATMENT IMAGING TREATMENT PLANNING
DAILY IMAGINGIMAGE REGISTRATION& PLAN ADAPTATION
TREATMENT DELIVERY
4D-CT + 4D-MRI
4D-MRI
6
1
75
4
83
2 9
10
MID-POSITION STRATEGY ON THE MR-LINAC
4D-MRI Daily4D-MRI
4D-CT
6
1
75
4
83
2 9
10
Simple Dose Shift
Simulation Pre-beam
6
1
75
4
83
2 9
10
10
midP-CT
midP-MRI
SELF-GATED 4D-MRI
DATA ACQUISITIONData from arbitrary respiratory phases
RESPIRATORY SIGNAL
DATA SORTINGTo amplitude or phase
11
2D SEQUENCE 3D SEQUENCECartesian Radial
K-space
4D-MRI ACQUISITION: 2D VS 3D
12
Acquisition 2D 3DSNR (spatialresolution)
- +
Acquisition / reconstruction time
~minutes (3-15min) <minute~hours / dedicated reconstruction server
Pulse sequences Standard TailoredContrasts T1 or T2 - weighted T1-weighted
13
2D-BASED 4D-MRI
2D Orientation Sagittal Axial CoronalCC-motion In-plane Out-of-plane In-planeFOV coverage / acquisition time
- + +
Axial
Sagittal
Coronal
FOV: 25 slices of 5mm, Acq time: ~4min
AXIAL/CORONAL 4D-MRI
RESULTS: SELF-SORTING SIGNAL ANALYSIS
15
Table1: Image‐based self‐sorting signal vs navigator signal resultsImS vs NavS Correlation 0.95 – 0.97 (range)RMSD (mm) 1.39 – 2.13 (range)Time difference inhale positons (s) 0.06 ± 0.13 (mean ± SD)Bin difference for amplitude‐binning 0.41 ± 0.64 (mean ± SD)Bin difference for phase‐binning 0.23 ± 0.47 (mean ± SD)
van de Lindt et al (2018), IJROBP
16
RESULTS: AMPLITUDE VS PHASE SORTING
van de Lindt et al (2018), IJROBP
RESULTS: TUMOR VISIBILITY
17
Lesion: 15mm
van de Lindt et al (2018), IJROBP
MR-LINAC: DAILY PLAN ADAPTATION IS A MUST
Adapted PLAN
Pre-beam
6
1
75
4
83
2 9
10
V
4D-MRI
Simple Dose Shift
Optimization
Contouring
InitialPLAN
midPMRI
Simulation
midPCT
MATERIALS & METHODS: EXPERIMENTAL VALIDATION
4D-MRI1• Unity MR-Linac• Coronal multi-2D TSE
TREATMENT PLANNING• Monaco v5.19• 10-beam step-and-shoot IMRT• 3x20 Gy
PHANTOM MOTION• Periodic CC motion
• A =15 mm• T=4 s
• CC baseline shifts• 0, 5, 10 and 15 mm
1 TN van de Lindt et al., IJROBP., 2018
RESULTS
EPID MRI FilmBaseline shift
(mm)MidP(mm)
Cine (mm)
4D(mm)
Residual CC-shift (mm)
γ pass rate(%)
5 5.0 5.3 3.8 1.6 – 1.8 86 – 9110 9.8 10.6 10.0 0.2 – 0.8 97 – 9915 15.0 15.5 14.3 -0.1 – 0.8 93 – 99
Δ Mean±SD 0.1±0.1 -0.5±0.1 0.6±0.6 0.8±0.8 94±5
GEOMETRIC AND DOSIMETRIC RESULTS
Gamma analysis: 3%/2mm
NEXT FRONTIER: REAL-TIME ADAPTATIONS
21
Motion variability not captured by 4d imaging
Fast et al (2018), IJROBP
CINE MRI + AUTO-CONTOURING
22
Dice1.0
0.8
0.6
0.4
0.2
0.0
TMA
DIR
PCN
N SIFT
Centroid dist. [mm]
TMA
DIR
PCN
N SIFT0
10
20
Fast & Eiben et al (2017), R&O
• 6 patients
• 22 image series: sequence (bSSFE vs GRE) + image orientation
• Dice: 0.92 (median)
• Centroid distance: 1.5 mm (median)
• Multi-template matching (MTM) and deformable image registration (DIR) performed best
TUMOUR TRAILING
REGULAR MOTION• High-frequency motion
components• Accounted for in mid-position
planning strategy
IRREGULAR MOTION• Low frequency motion components
(slow baseline drifts)• Following mid-position changes
during treatment Tumour trailing#
23
midPRespiratory motion
Time
#Trofimov et al (2008), Med. Phys. 35 / George et al (2008), Med. Phys. 35
MR-LINAC: FEASIBILITY OF TRAILING
• 3x20 Gy liver SBRT
• Simulated delivery to midP CT (full consideration of interplay)
• Baseline drift: 0.5 mm/min (CC) & 0.25 mm/min (AP)
• Isocentre-shift ‡ dose reconstrucion
• Trailing vs Conventional vs Planned
‡Poulsen et al (2012), Med. Phys. / Fast et al (2018), IJROBP
Conventional - planned
Trailing - planned+6 Gy
-6 Gy
24
MR-LINAC: TRAILING PROOF-OF-PRINCIPLE
25
Motion monitoring• Sagittal b-FFE cine-MRI• 1x1x10 mm3
• 3 Hz
Plan adaptation• Triggers for motion ≥1
mm• Adapts MLC to target in
beam’s-eye-view
Trailing• Starts with initial plan• Pauses linac when needed• Transfers adapted plan via
iCom interface
RESULTS: DOSE AND GAMMA-DISTRIBUTION
26Gamma analysis: 3%/2mm
Pre‐treatment imaging Treatment‐planning
0 20 40 60 800
0.2
0.4
0.6
0.8
1
Dose (Gy)
Volu
me
(%)
In‐room imaging Image registration &correction
Treatment delivery (~30/35 fx)
The (adaptive) Radiotherapy Process
Intervention (ART)
Treatment assessment EPID dosimetryDecision rulesTraffic‐light protocol
Introduction – current practice
• EPID: traffic‐light for dose differences with planned>> but what does it mean in the anatomy?
• IGRT: Decision rules & traffic‐light protocol>> we see deformations, what does it mean?– Dual registration Lung– Multi‐clipbox HNC
• Ad hoc assessment: editing the pCT for weightloss/deformations in Pinnacle …
Example lung
• Lung Cancer• 60 Gy, 30 fxs• Previously
irradiated: cord dose < 20 Gy NTD (a/b=2)
Example HNC
• Nasopharynx• SIB 70 Gy, 35 fxs• Large PTV• Many OARs:
N. Opticus, Chiasm, Brainstem
• mROI with weights 1/0
What these examples show
Mental exercise…• Isodose lines: is that what we delivered?• Deformations: how does that change CTV/OAR dose?• History: Systematic? Random? Trend?• Timing: How much can we still gain with replanning?• Upon replanning: what about dose
already delivered?
We see geometry, we must think dose…
>> HEADACHE!
Alternative: dose accumulation
alternative: dose accumulation
compute the total delivered dose in the patient while accounting fordosimetrical and geometrical discrepancies
2 important steps:
• Dose distribution delivered is different than planned
• Dose ends up at different location than planned
Dose accumulation
Step 1: Dose distribution delivered is different than planned
Planned dose in pCT planned/recalculated dose in CBCT
Dose accumulation
Step 2: Dose ends up at different location than planned
Deformation & anatomy differences
CBCT warping
Dose accumulation
Step 2: Dose ends up at different location than planned
Deformation & anatomy differences
dose warping
Dose accumulation
After dose warping: common reference
planned dose warped dosedose difference
Planned dose Accumulated dose Difference dose
Dose accumulation
DVH: accumulated dose vs planned dose
BRAINSTEM
SPINAL_CORD
PAROTID
CTVCTVsd
Dose accumulation: DVH differences
LARYNX — planned dose
‐‐‐ accumulated dose
ORAL VACITY
Challenges: prediction of final dose
ART @ fx 14
Boost plan
Challenges: DIR with limited Image Quality• HNC: OK…• Lung: WIP…
CBCT MRExample image quality rectum
Courtesy Sander Uilkema & Lisa Hartgring
46
DOSE ACCUMULATION: ENERGY-MASS TRANSFER
Li et al. (2014), Med. Phys. / Ziegenhein et al (2018), Sci. Rep.
SUMMARY & OUTLOOK
• MRI-guidance provides excellent soft-tissue contrast
• Novel 4D-MRI radiotherapy solutions are being developed
• Margin reduction / active motion management will provide scope to allow for re-irradiations
• Away with the surrogate reality: start looking at dose in anatomydaily